tjthurman / bahz Goto Github PK

Try the Cmd+Shift+A command.

Would Stan still fit a model, given a negative width as an initialization value?

In particular, need it to help proide initial values, as it doesn't seem to be working with random generation. Stan randomly generates initial values "between -2 and 2 on the unconstrained support". Not really sure what that means, but seems bad given that a bunch of my parameters can't be random numbers.

Ideally, it would just draw a random number from the prior distribution. That might be a little tough to implement, could take some matching to the yaml. But, if I could make it generate the expression for one chain, I could just copy that expression somehow for multiple chains.

Functions that will turn data frames into the appropriate lists for input into stan

Might be too much, but could I create a little function, a la precis in rethinking, that basically does the same summarizing but returns a data frame?

Do we need to create all 20 models? or just the ones the user wants?

A function that plots a cline, given cline parameters and an x range? Wouldn't be too hard to build off the gen_cline_equation function.

Specifically, say what column(s) are missing when you use fit_geno_cline

Might ease initialization in the case of poorly specified center priors.

Might ease initialization. Could add the mean value back to the results in the chain, to preserve the user-supplied values. Need to see if Stan would work with the center = 0 check removed.

Some ideas to unify fxn names, and prep for future pheno cline implementation:

prepping functions:

load_cline_data -> prep_geno_data DONE
make_init_list -> prep_init_list DONE
make_prior_list -> prep_prior_list DONE

Main user functions:

fit_cline -> fit_geno_cline DONE
sim_data_from_cline -> sim_geno_cline DONE

Could I allow f to be <0, for an excess of heterozygotes?. Only issue would be if it tried to make a genotype category be more than 1 of less than 0

When running the package checks in RStudio, the column names within the dplyr pipeline always come up as undefined global variables. See if there's a way to fix that.

I've fixed the function-specific ones for sim_data from cline. Now its just the dplyr global variables, in a few functions, that it gets mad about.